Simulation run on this device. Scenario values are not sent anywhere — the engine is fully local.
research studio · pro
Paper-to-Code Repro Checker
Score a paper's reproducibility surface: code, seeds, splits, configs.
Is this result reproducible from what's published?
Engine 1.0.0 · studio
- Reproducibility
- 0/10
- Signals found
- 0/5
- Paper length
- 0 words
- mediumMissing: Code / weights released
no code/weights availability signal
- mediumMissing: Hyperparameters
learning rate / batch size / optimizer not specified
- mediumMissing: Dataset & splits
dataset identity or train/val/test split unclear
- mediumMissing: Seeds & environment
seeds / hardware / software versions absent
- mediumMissing: Eval protocol
metric definition or baselines missing
Each missing dimension is a guess a reproducer would have to make; the first one is usually the highest-value fix.
Method
- Scores five reproducibility dimensions 0–2 from explicit signals in the text: code/weights, hyperparameters, dataset & splits, seeds/environment, and eval protocol.
- A missing dimension is a guess a reproducer would have to make — the score is a checklist, not a promise.